Analyzing QB projected vs actual win %

[Kjmass1]

I thought this article had some great analytics on QB performance.

A couple takeaways-

- Manning overall has a pretty solid winning performance in all games
- Brady overperforms in games they are projected to lose, but falls back to historical average in games they are supposed to win
- Luck kills it on games they have no business winning
- Stafford is the master of losing games they should win


http://fivethirtyeight.com/features/skeptical-football-special-midterm-edition/

 

[Saint Pat]

Left side of the straight dashed line = good, right side of the straight dashed line = bad. In conclusion, why is Fitzpatrick still in the NFL?

 

[RobertWeathers]

Fitzpatrick is an absolute bum.

Dirty Sanchez actually does ok. Even though he was carried by good defenses for half of his career.

Eli's at the tail-end is interesting.

Everyone needs to look out once Luck has a defense supporting him.

 

[Saint Pat]

Look at Brady's curve! At zero percent projected win percentage, the actual win percentage is about 10% greater than zero ----------> CLUTCH!
With Tom Brady you always have a chance.

 

[PATS16N0]

I see colors and lines.

 

[Kjmass1]

Look at Brady's curve! At zero percent projected win percentage, the actual win percentage is about 10% greater than zero ----------> CLUTCH!
With Tom Brady you always have a chance.

That is pretty incredible, when you consider the sample size of how many games this is averaged over.

 

[Saint Pat]

I see colors and lines.

Left side of the straight dashed line = good, right side of the straight dashed line = bad. In conclusion, why is Fitzpatrick still in the NFL?

 

[Canuck]

Some thoughts based on above graph: I'm surprised by Rodger's graph, it shows a consistent slightly above average winning % . Hard to say what's going to happen in the future but Luck looks like he is going to be compared to Brady and Peyton. Dalton shows he is not clutch and loses games he should have won. Big Ben's looks like what I expected, solid and above average. Carson Palmer! How in the heck is Arizona #1 in the NFL at 7-1. No surprises with Brees, a solid looking graph. Looking at Eli's graph and I am starting to doubt how valid the data making up the graph is worth. Flacco may just be worth the money they paid him. Sanchez not surprised, average at best. Stafford just wow, wins games he should lose and loses games he should win. Peyton best looking graph in the regular season. River's a little surprised that he is avg to below avg on games he should win. Fitzpatrick shouldn't ever be a starter with a graph like that. Brady is probably one of the best at winning games he should lose. When your expected to win 75% of your games a few stinkers like Kansas City can flatten your avg at the top. Romo has the flip flops, does really well in games he should lose and worse in games he should win

 

[Bobsyouruncle]

Somebody expected Brady to lose a game?

 

[CaffeineMan]

Yeah, I ran across these graphs today. Pretty interesting. One caveat on these, I think: like Football Outsiders, you can't really separate the player from their offense, because they are looking at the results of the plays, not scouting the plays individually. So Eli really means "Eli in the Giants offense". And I'm not sure these are adjusted for opponent, but I might be wrong there.

Still, I find stuff like this pretty cool. And some of the graphs are almost hilarious (Stafford).

 

[ALP]

so this isnt based solely on projections at start of the game, but at any point during the game? so if we're down by whatever, and our win % is at 10%, and we pull through, its counted as a win?

im no stats guy, but i think theres a LOT of in depth stuff that needs to be explained with these graphs/mentality...

 

[wolverinejoe80]

interesting stuff, but at the end of the day, only thing that matters is Brady is the greatest winner in the NFL.

just get that 4th ring, so we can all die happy.

 

[CaffeineMan]

so this isnt based solely on projections at start of the game, but at any point during the game? so if we're down by whatever, and our win % is at 10%, and we pull through, its counted as a win?

im no stats guy, but i think theres a LOT of in depth stuff that needs to be explained with these graphs/mentality...

Yeah, I agree. The chart was listed in a section they have entitled "Experimental Chart of the Week" or something like that. Not a full-blown statistical system like DVOA/DYAR, but more of a fun quick take. Advanted stats in football remind me of that saying that goes something like: "Democracy is a terrible system of government… but all the other ones are worse."

 

[Ochmed Jones]

This is the first real trap game of the year.

 

[lurker1965]

Ok. Time to call BS on the base stat. Win over projected based on what? Yes, you, or they as in a generic you or anyone, can produce a 90 page break down of all the factors. However anyone with any sense at all can quibble with any or all of the factors.

For example: The opponent had a #1 in yards, but only #7 defense in points therefore had a 7 percent chance of losing on a Thursday in the rain when the temp was over 50 degrees blah, blah, blah, ad infinitum. But it ignores their +8 turnover differential and #8 in scoring on special teams, yada, yada, yada, ad infinitum.

The more complex the stat, the more this bickering with it's base is invited. A stat based upon a stat based upon stats is inherently ridiculous.

 

[Canuck]

I went back and read the whole article. The author was doing the typical mid season awards and using his unique stat system to come up with the awards. The funny thing his system says the Aaron Rodgers was the mid season MvP but he ended up giving the award to Manning. The whole article was just having fun with stats. I don't think he meant for us to take this very seriously.

 

[primetime]

so this isnt based solely on projections at start of the game, but at any point during the game? so if we're down by whatever, and our win % is at 10%, and we pull through, its counted as a win?

im no stats guy, but i think theres a LOT of in depth stuff that needs to be explained with these graphs/mentality...

It's similar to a weather forecast. Based on time remaining and the score (and possibly other factors, like down and distance), historically what is the likelihood that a typical team wins a typical game? The major difference with the weather forecast is that it brings the human component in, which allows talent (or human error) to overcome the odds in addition to plain randomness (or luck, if we want to call it that). So this is showing who consistently beats that win probability at a given time.

Of course, this doesn't take into account a defense's or other players' role in helping win, and it rewards miracle, heart attack-inducing wins over clinical ones. Brady's probably getting a ton of credit for, say, the wins against Denver, the Saints, and Browns last season.

For example, here's the win probability chart from last week's game (you can monitor these live at http://live.advancednflstats.com/weekly.php?gameid2=56300&week=9)